distributed system behavior modeling of urban systems with …austin/ence688r.d/lecture... ·...

Problem Statement Related Work Contributions Semantic Modeling Case Studies 1 and 2 Conclusions

Distributed System Behavior Modeling of UrbanSystems with Ontologies, Rules and

Many-to-Many Association Relationships

Maria Coelho, Mark A. Austin, Mark Blackburn

University of Maryland, Stevens Institute of Technology

[email protected], [email protected], [email protected]

Presentation at ICONS 2017, Venice, Italy

April 22, 2017


Overview

1 Problem Statement

2 Related Work

3 Contributions

4 Semantic Modeling

5 Case Studies 1 and 2

6 Conclusions


Problem Statement


Interdependent Urban Networks

Networks are heterogeneous,interwoven, dynamic.

Disciplines want to operateindependently in theirdomain.

Achieving target levels ofperformance and correctnessof functionality requiresdisciplines to coordinateactivities at key points insystem operation.

Services

Waterway Network

Transportation Network

Information andCommunications

Emergency


Cascading Failures

Disturbance in one system canimpact other networks inunexpected, undesirable and costlyways.

Often, infrastructure managementsystems do not allow manager ofone system to access operationsand conditions of another system.

Decision making is complicated bypresence of newfound systeminteractions, incomplete knowledgeof system state, and break downsof communication among urbannetworks.


Long-Term Project Objective: City Operating System

Urban

monitor

monitor

actionsSpace−time terrain

interacting with service infrastructures

Environmental Processes

processes

Monitoring

Evaluation

Reasoning

Relief Actions

City Operating System


Short-Term Project Objective: Behavior Modeling

Ability to model behavior of city-domain processes, andinteractions among distributed system behaviors within a city.

Transportation System

Transportation Domain

Metro System Routes

Bus RoutesUrban Business

Business / Work Domain

MediatorBusiness − Trans.

Mediator

Flows of: information, goods, energy.

Flows of: information, goods, energy.

goods, energy.Flows of: information, Flows of: information,

Physical Infrastructure Domain

Power Network

OptaPlanner: Real−Time Network Control and Planning for System Recovery

Infrastructure − BusinessGovernmentDepartment

−− Behavior control−− Resilience assessment−− Planning for receovery



Utility Network

goods, energy.

Physical System Business System


Benefits of Behavior Modeling

Allows decision makers to understand:

How failure in one network will impact other networks.

What parts of a system are most vulnerable.

Allows decision makers to assess:

Sensitivity of systems to model parameter choices.

Influence of resource constraints.

Potential emergent interactions among systems.


Solution Approach


Systems of Systems Perspective

Cities are System of Systems.

City subsystems may have a preference to operating asindependently as possible from the other subsystems.

Strategic collaboration among subsystems is often needed toeither avoid cascading failures across systems and/or recoverfrom a loss of functionality.

AirportHome Taxi Airport Airplane Taxi Destination


Observing Urban Behavior

Traffic LightRoad Network

Pedestrian

Automobiles

Traffic Control


Ontologies, Rules, and Reasoning Mechanisms

ReasonerProperties

Instances

DataRequirementIndividual

verify

Textual Requirementsdefine

Classes

Relationships

Ontologies and ModelsDesign Rules and Reasoner

Design Rules

Engineering Model

System Structure

System Behavior

Remarks

System structures are modeled as networksand composite hierarchiesof components.

differential equations.represented by partial

state machines.modeled with finite Discrete behavior will be

associated with components.Behaviors will be

a c d

b

Continuous behavior will be


Ontologies, Rules, and Reasoning Mechanisms

load Abstraction

Road network model

Automobile model

Traffic control model

Traffic light model

Engineering Model and Data

RoadNetwork.owl

Automobile.owl

TrafficLight.owl

TrafficControl.owl

RoadNetwork.rules

Automobile.rules

TrafficLight.rules

TrafficControl.rules

Domain Rules Ontology Classes & Properties

Pedestrian.rules Pedestrian.owl Pedestrian model

Semantic Graphsload

data

events

Spatial.owl

Spatial.rules

PhysicalQuantity.owlTime.owl

Time.rules PhysicalQuantity.rules

Cross−cutting (fundamental) Ontologies & Rules

load

load

graph transformation

Reasoner


Related Work


Glassbox Simulation Engine


Graphs, Cellular Automata, and Ontologies

Numerous researchers have studied the topology of urbanenvironments from a graph theoretic standpoint.

Other studies capture the temporal dynamics of cities withcellular automata, agent-based models, and fractals.

Extensive studies have been conducted on the development ofontologies for the geographic information sector.

Researchers have proposed so called smart city ontologies.

A notable effort in the direction of ontologies developedalongside rules is the DogOnt ontology model.


Contributions


Contributions

Framework for modeling concurrent, directed communicationbetween all entities composing a system.

Mediator

Mediator−Enabled Communication

System−to−System Communication

Mechanisms for incorporating notions of space and time in thereasoning process.


Semantic Modeling


Introduction to the Semantic Web

Extension to the World Wide Web

Allows machines to access and share information.

Relies on technical infrastructure below.


Working with Jena and Jena Rules


Distributed Behavior Modeling

listener

Semantic Model: Domain 1 Semantic Model: Domain 2

Rules for domain 1 Rules for domain 2

AbstractOntologyModel<< abstract >>

importimport

listens for ModelChange events

message input

message input

AbstractOntologyInterface<< abstract >>

message passing

MediatorInterface: Domain 1 Interface: Domain 2

message passing


Case Study 1


Family-School System Dynamics

Report

Family Graph

Model Model

listen

Family Interface

Family Domain

import

Reasoner

family rules

family − schoolinteraction rules

School System DomainMediator Domain

school systemrules

ReasonerReport

Enrollment Enrollment

import

Graph Model

School System

listen

School System

Interface ModelMediator

import


Framework for Communication

Family Domain

Elementary School

Middle School

High School

MediatorFamily B

Family C

Family A

School Domain


Generation of Semantic Models


Family XML Datafile

<?xml version="1.0" encoding="UTF-8"?>

<FamilyModel author="Maria Coelho" date="2017" source="UMD">

<Family>

<attribute text="FamilyName" value="Austin"/>

<attribute text="Address" value="6242 Heather Glen Way, Clarksville, MD 21029"/>

<Person>

<attribute text="Type" value="Male"/>

<attribute text="FirstName" value="Mark"/>

<attribute text="MiddleName" value="William"/>

<attribute text="LastName" value="Austin"/>

<attribute text="BirthDate" value="1704-06-10"/>

<attribute text="Weight" value="170.0"/>

<attribute text="Citizenship" value="New Zealand"/>

<attribute text="SocialSecurity" value="111"/>

</Person>

<Person>

... description of other Austin family members ....

</Person>

</Family>

<Family>

<attribute text="FamilyName" value="Jones"/>

<attribute text="Address" value="5807 Laurel Leaves Ln, Clarksville, MD 21029"/>

<Person>

... description of Jones family members....

</Person>

</Family>

</FamilyModel>


School XML Datafile

<?xml version="1.0" encoding="UTF-8"?>

<SchoolSystemModel author="Maria Coelho" date="2017" source="UMD">

<School>

<attribute text="Type" value="High School"/>

<attribute text="Name" value="River Hill High School"/>

<attribute text="Grade" value="Grade09"/>




<attribute text="Report Period Start Time" value="2016-09-01T00:00:00"/>

<attribute text="Report Period End Time" value="2020-10-20T00:00:00"/>

</School>

<School>

... description of Clarksville Middle School ...

</School>

<School>

... description of Pointers Run Elementary School ...

</School>

</SchoolSystemModel>


Family Ontology


School Ontology


Family Rules - A Sample

@prefix af: <http://austin.org/family#>.

@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>.

// Rule 01: Propagate class hierarchy relationships ....

// Rule 02: Family rules ....

// Rule 03: Identify a person who is also a student ...

[ Student: (?x rdf:type af:Person) (?x af:hasAge ?y)

greaterThan(?y, 4)

lessThan(?y, 18) -> (?x rdf:type af:Student) ]

[ UpdateStudent: (?x rdf:type af:Student) (?x af:hasBirthDate ?y)

getAge(?y,?b) ge(?b, 18) -> remove(0) ]

// Rule 04: Compute and store the age of a person ....

[ GetAge: (?x rdf:type af:Person) (?x af:hasBirthDate ?y)

getAge(?y,?z) -> (?x af:hasAge ?z) ]

[ UpdateAge: (?a rdf:type af:Person) (?a af:hasBirthDate ?b)

(?a af:hasAge ?c) getAge(?b,?d)

notEqual(?c, ?d) -> remove(2) (?a af:hasAge ?d) ]


Behavior Modeling Use Cases


Case Study 2


Family-School-Urban-Geography System Dynamics

Clarkesville Elementary School

Clarksville Middle School

Riverhill High School

Pointers Run Elementary School

Clarksville Elementary School

Riverhill High School

Poinrers Run Elementary School

Clarksville Middle School

School Zone Boundary for

Clarksville Elementary

School Zone Boundary for

Pointers Run


Extensions to Ontologies and Rules

Certain ontology properties are added to the framework inorder to allow modeling spatial behavior (e.g.livesInSchoolZoneOf, isEligibleForSchoolBus)

Additional rule determines whether or not a person is eligibleto the school bus service

Additonal rule only allows students to enroll when they livewithin the school zone jurisdiction.

Graph transformations in the school system model can nowoccur due not only to input or time, but also space.


Accessing Spatial Data from OpenStreetMap


Conclusions

and

Future Work


Conclusion and Future Work

Project focused on design and preliminary implementation ofmessage passing infrastructure needed to supportcommunication in many-to many association relationshipsconnecting domain-specic networks.

Long-term objective is to build upon family-school distributedbehavior model and create models of distributed behavior ofurban infrastructure multi level systems, and simulatecascading system failures that occur due to extreme externalevents.

Domain interfaces have been assumed to be homogeneous,but will not always be the case.

Need for new approaches to the construction and operation ofmessage passing mechanisms.


Future Work

among Networked Domains.

Mechanisms for Message Transmisson and Processing in Apache Camel.

Message

Endpoint

Cha

nnel

Neworked Domain 2Networked Domain 1

Message

Endpoint

Cha

nnel

Cha

nnel

Message

Endpoint

Networked Domain 3

Distributed System Behavior Modeling

ImportIntelligent Routing of Messages

Platform Infrastructure for

Message−based Routing

Content−based Routing

Message−based Translation

Message Filtering


Extra Slides


Model-Based Systems Engineering

Understanding the relationships among the networks and theircombined behaviors can be very challenging.

City systems are being upgraded from industrial-age capabilityto information-age capability.

Challenges can be mitigated through the systematicapplication of model-based systems engineering (MBSE)procedures.

State-of-the-art MBSE procedures fall short is in thesystematic consideration of interactions among manyconcurrent behaviors.


Solution Approach: Systems of Systems Perspective

A System of Systems is a collection of independentlyoperational systems which have been glued together toachieve further emergent properties.

The component systems operate for their own purposes ratherthan the purposes of the combined system.

Yet, they also function to resolve the purposes of the wholewhich are generally unachievable by the individual systemsacting independently.

The system of systems will change over time as constituentsystem are replaced.


Working with RDF

RDF is a graph-based assertional data model for describingthe relationships between objects and classes.

Assertions are transformed into RDF triples consisting of asubject, a predicate and an object.

A set of related triples constitute an RDF graph


Working with RDF


Working with OWL

RDF is unable to capture existence,cardinality, localized range,domain constraints, transitivity, inverse or symmetricalproperties.

OWL was developed to address the weaknesses of RDF.

The additional capabilities allow ontological systems to usereasoning to infer new triples from existing ones.


Working with OWL


Working with OWL

// Define Classes ...

<owl:Class rdf:about="http://example.org/monaLisa#Painting">

</owl:Class>

<owl:Class rdf:about="http://example.org/monaLisa#Person">

</owl:Class>

<owl:Class rdf:about="http://example.org/monaLisa#Museum">

</owl:Class>

// Define Datatype Properties ...

<owl:DatatypeProperty rdf:about="http://example.org/monaLisa#hasType">

<rdfs:domain rdf:resource="http://example.org/monaLisa#Painting"/>

<rdfs:range rdf:resource="&xsd;string"/>

</owl:DatatypeProperty>

<owl:DatatypeProperty rdf:about="http://example.org/monaLisa#hasCompletionDate">


<rdfs:range rdf:resource="&xsd;date"/>

</owl:DatatypeProperty>


Working with OWL

// Define Object Properties ...

<owl:ObjectProperty rdf:about="http://example.org/monaLisa#hasCreator">


<rdfs:range rdf:resource="http://example.org/monaLisa#Person"/>

</owl:ObjectProperty>

<owl:ObjectProperty rdf:about="http://example.org/monaLisa#hasLocation">


<rdfs:range rdf:resource="http://example.org/monaLisa#Museum"/>

</owl:ObjectProperty>


Working with Jena and Jena Rules

Apache Jena is an open source Java framework for buildingSemantic Web and linked data applications.

Jena provides APIs for developing code that handles RDF,RDFS, OWL and SPARQL.

Jena inference subsystem is designed to allow a range ofinference engines or reasoners to be plugged into Jena (e.g.Jena Rules).

Jena Rules use facts and assertions described in OWL to inferadditional facts from instance data and class descriptions.

Such inferences result in structural transformations to thesemantic graph model.


System Modeling Assumptions

All of the models execute under a single continuous thread ofcomputation, with the only interaction among domains beingexchange of messages.

Delays in communication between domains are ignored.

Behavior models are deterministic; uncertainties in behaviorare ignored.

Support for fault-tolerant communication among domains isignored. We do, however, send confirmation messages back tothe sender.


Family Rules

@prefix af: <http://austin.org/family#>.



[ rdfs01: (?x rdfs:subClassOf ?y), notEqual(?x,?y),

(?a rdf:type ?x) -> (?a rdf:type ?y)]

// Rule 02: Family rules ....

[ Family: (?x rdf:type af:Family) (?x af:hasFamilyMember ?y) ->

(?y af:belongsToFamily ?x) ]

// Rule 02: Identify a person who is also a child ...

[ Child: (?x rdf:type af:Person) (?x af:hasAge ?y)

lessThan(?y, 18) -> (?x rdf:type af:Child) ]

[ UpdateChild: (?x rdf:type af:Child) (?x af:hasBirthDate ?y)


// Rule 03: Identify a person who is also a student ...

[ Student: (?x rdf:type af:Person) (?x af:hasAge ?y)

greaterThan(?y, 4) lessThan(?y, 18) -> (?x rdf:type af:Student) ]

[ UpdateStudent: (?x rdf:type af:Student) (?x af:hasBirthDate ?y)


// Rule 04: Compute and store the age of a person ....

[ GetAge: (?x rdf:type af:Person) (?x af:hasBirthDate ?y)

getAge(?y,?z) -> (?x af:hasAge ?z) ]

[ UpdateAge: (?a rdf:type af:Person) (?a af:hasBirthDate ?b) (?a af:hasAge ?c)

getAge(?b,?d) notEqual(?c, ?d) -> remove(2) (?a af:hasAge ?d) ]

// Rule 05: Set father-son and father-daughter relationships ...

[ SetFather01: (?f rdf:type af:Male) (?f af:hasSon ?s)-> (?s af:hasFather ?f)]

[ SetFather02: (?f rdf:type af:Male) (?f af:hasDaughter ?s)-> (?s af:hasFather ?f)]


School Rules

@prefix af: <http://austin.org/school#>.



[ rdfs01: (?x rdfs:subClassOf ?y), notEqual(?x,?y),

(?a rdf:type ?x) -> (?a rdf:type ?y)]

// Rules 02: Elementary school rules ...

[ EnterElementarySchool: (?x rdf:type af:Student) (?y rdf:type af:ElementarySchool)

(?x af:hasBirthDate ?a) getAge(?a,?b) ge(?b, 6) le(?b, 10) ->

(?x af:attendsElementarySchool af:True) (?y af:hasStudent ?x)]

[ LeaveElementarySchool: (?x rdf:type af:Student) (?x af:hasBirthDate ?a)

(?x af:attendsElementarySchool af:True) (?y af:hasStudent ?x)

getAge(?a,?b) ge(?b, 10) -> remove(2) ]

[ GradeOne: (?x rdf:type af:Student) (?x af:hasBirthDate ?a)

getAge(?a,?b) equal(?b, 6) -> (?x af:isInGrade af:Grade01) ]

... Rules for Grades 2 through 5 removed ...

// Rules 03: Middle school rules ...

... Middle school rules removed ...

// Rules 04: High school rules ...

...High school rules removed ...

// Rules 05: If today is report period, send school report ....

[ GenerateReport: (?x rdf:type af:Event) (?y rdf:type af:Student)

(?z rdf:type af:School) (?z af:hasStudent ?y) (?x af:hasStartTime ?t1)

(?x af:hasEndTime ?t2) getToday(?t3) lessThan(?t3,?t2)

greaterThan(?t3,?t1) -> (?y af:hasReport af:True) ]


Accessing XML Data from Data Models

distributed system behavior modeling of urban systems with …austin/ence688r.d/lecture... ·...

Documents